Analysis of Emission Data from Global Commercial Aviation: 2004 and 2006

Atmos. Chem. Phys., 10, 6391–6408, 2010 www.atmos-chem-phys.net/10/6391/2010/ Atmospheric doi:10.5194/acp-10-6391-2010 Chemistry © Author(s) 2010. CC Attribution 3.0 License. and Physics

Analysis of emission data from global commercial aviation: 2004 and 2006

J. T. Wilkerson1, M. Z. Jacobson1, A. Malwitz2, S. Balasubramanian2, R. Wayson2, G. Fleming2, A. D. Naiman3, and S. K. Lele3 1Civil and Environmental Engineering, Stanford University, Stanford CA, USA 2Volpe National Transportation Systems Center, US Department of Transportation, Cambridge, MA, USA 3Aeronautics and Astronautics, Stanford University, Stanford CA, USA Received: 21 December 2009 – Published in Atmos. Chem. Phys. Discuss.: 5 February 2010 Revised: 16 June 2010 – Accepted: 18 June 2010 – Published: 14 July 2010

Abstract. The global commercial aircraft fleet in 2006 flew ally in 2006, more than half over three regions: the United 31.26 million flights, burned 188.20 million metric tons of States (25.5%), Europe (14.6), and East Asia (11.1). De- fuel, and covered 38.68 billion kilometers. This activity spite receiving less than one percent of global emissions, the emitted substantial amounts of fossil-fuel combustion prod- Arctic receives a uniformly dispersed concentration of emis- ucts within the upper troposphere and lower stratosphere sions with 95.2% released at altitude where they have longer that affect atmospheric composition and climate. The emis- residence time than surface emissions. Finally, 85.2% of all sions products, such as carbon monoxide, carbon dioxide, flights by number in 2006 were short-haul missions, yet those oxides of nitrogen, sulfur compounds, and particulate mat- flights were responsible for only 39.7% of total carbon from ter, are not emitted uniformly over the Earth, so understand- CO2. The following is a summary of these data which illus- ing the temporal and spatial distributions is important for trates the global and regional aviation emissions footprints modeling aviation’s climate impacts. Global commercial for 2004 and 2006, and provides temporal and spatial distri- aircraft emission data for 2004 and 2006, provided by the bution statistics. Volpe National Transportation Systems Center, were computed using the Federal Aviation Administration’s Aviation Environmental Design Tool (AEDT). Continuous improve- 1 Introduction ment in methodologies, including changes in AEDT’s horizontal track methodologies, and an increase in availability This study provides an analysis of aviation emissions data for of data make some differences between the 2004 and 2006 global commercial aircraft flights in 2004 and 2006. Previous inventories incomparable. Furthermore, the 2004 inventory aircraft emissions inventories have shown that most activity contained a significant over-count due to an imperfect data occurs in the Northern Hemisphere mid-latitudes (Baughcum merge and daylight savings error. As a result, the 2006 emis- et al., 1996a, b; Sutkus Jr. et al., 2001; Eyers et al., 2005; sions inventory is considered more representative of actual Kim et al., 2005c, 2007). Simply scaling data from earlier flight activity. Here, we analyze both 2004 and 2006 emis- studies by projected global industry growth rate of 5–7% sions, focusing on the latter, and provide corrected totals for per annum (ICAO, 2007) may not provide emission trends 2004. Analysis of 2006 flight data shows that 92.5% of fuel that are representative of geographically varying growth in was burned in the Northern Hemisphere, 69.0% between 30N the aviation sector. For example, India reported an increase and 60N latitudes, and 74.6% was burned above 7 km. This in its domestic aviation activity by 41% from 2005 to 2006 activity led to 162.25 Tg of carbon from CO2 emitted glob- (MOCA, 2007). As such, it is necessary to update and analyze aircraft emission inventories for use in atmospheric models and policy studies. Correspondence to: J. T. Wilkerson ([email protected])

Published by Copernicus Publications on behalf of the European Geosciences Union. 6392 J. T. Wilkerson et al.: Analysis of emission data from global commercial aviation

Data for this analysis were provided by the Volpe National 2 Methodology Transportation Systems Center. Here, we provide a brief review of the processes and methods used by Volpe to as- 2.1 Global inventory source and description semble the data and provide emission summary statistics for carbon dioxide (CO2), nitrogen oxides (NOx), sulfur com- The United States (US) Federal Aviation Administration pounds (SOx), speciated organic gases, and speciated parti- (FAA), with support from Volpe, developed the Aviation En- cle components (black carbon, primary organic matter, and vironmental Design Tool (AEDT) (Roof et al., 2007). The sulfate) from commercial aircraft. We also analyze the verti- emission module of AEDT, the System for assessing Avia- cal profile of emissions, particularly in the upper troposphere tion’s Global Emissions (SAGE), was developed to provide and lower stratosphere (UTLS) and emissions in major re- a method for evaluating the effects of various policy, tech- gions of the world. We then examine the emissions over the nology, and operational scenarios on aircraft fuel use and Arctic, since it is a particularly sensitive region to climate emissions. The tool predicts aircraft fuel burn and emissions change. Finally, we compare results with those from previ- for all global commercial flights at high temporal resolution ous datasets. We do not examine general aviation or mili- (seconds to minutes) over each year, enabling single-flight tary aircraft emissions because such emissions are generally analysis scenarios, and airport, country, regional, and global not reported or reported separately and not provided. It has scenarios. The module dynamically models aircraft perfor- been estimated that the military contribution is in the range mance, fuel burn and emissions, capacity and delays at air- of 10–13% of total emissions (Eyers et al., 2005; Waitz et al., ports, and is also capable of providing forecast scenarios. 2005). The process for obtaining the flight-level emissions is briefly Aviation emission inventories are important inputs into at- summarized below but is described in much greater detail in mospheric models examining the climate and pollution ef- several publications (Kim et al., 2005a, b, c, 2007; Lee et al., fects of aircraft and other anthropogenic emission sources. 2005, 2007; Malwitz et al., 2005). For example, both inventories discussed here are currently The AEDT inventory provides four-dimensional (latitude, being used in companion studies to examine the effects of air- longitude, altitude, and time) flight trajectories using as much craft on global climate, atmospheric composition, contrails, real data as possible as described below. The data are col- and contrail-induced cirrus by treating each aircraft flight at lected by the Enhanced Traffic Management System (ETMS) the subgrid scale (Jacobson et al., 2010). It would not be at the Volpe Center, which serves as the hub of informa- possible to study the effects of subgrid-scale aircraft exhaust tion. ETMS receives a continuous flow of data from nu- plumes and contrails in a global model with an inventory that merous sources, including Terminal Radar Approach Control did not separate emissions by individual flights broken into Facilities (TRACON), individual airlines, Automated Radar spatial and temporal segments. Due to the large volume of Tracking Systems (ARTS), and Air Route Traffic Control data in such an inventory, it is important to evaluate and sum- Centers (ARTCC). ETMS is the FAA’s electronic recording marize the data in a meaningful way. We attempt to do this of flight position and flight plan information used for air traf- in the following sections. fic management. It captures every flight within coverage This paper presents emissions inventories from two years, of FAA radars, including scheduled, cargo, military (later 2004 and 2006, along with additional statistics for a cor- excluded), charter, and unscheduled flights. Unscheduled rected 2004 inventory. Differences between the original 2004 flights have been estimated to account for as much as 9% inventory and the 2006 inventory, due to changes in Volpe’s of flights globally; in the non-radar portions of AEDT data methodology, availability of radar data, emissions Indices unscheduled flights are included by scaling known flights (EI), and other factors, are discussed herein. Because of these (Kim et al., 2005b). ETMS also captures information about differences, comparisons between the original 2004 and 2006 every flight that files a flight plan, whether or not the air- data are limited to global methodology comparisons, and craft enters radar-controlled airspace. Radar coverage en- those data sets should not be compared directly with each compasses all of North America and parts of Western Eu- other to show spatial trends between the two years. How- rope, and records an estimated 50–60% of global commer- ever, corrected 2004 global emission totals based on the 2006 cial flights (Volpe, 2003). Additionally, in the 2006 inven- methodology, are presented. These numbers do allow a com- tory, data collected by EUROCONTROL’s Enhanced Tacti- parison of global trends between 2004 and 2006. The 2006 cal Flow Management System (ETFMS) were included. This dataset is considered the benchmark inventory. Below, we expanded AEDT’s total radar-provided schedule and flight address the 2006 data set and both the original and corrected data to include most of Europe. For the 2004 inventory data, 2004 inventories. European operations were represented primarily by the Of- ficial Airline Guide (OAG), where ETMS coverage did not exist (e.g., in Europe). When radar tracking is unavailable or information is incomplete for a particular flight, the tool refers to the OAG, which lists scheduled passenger flights by participating

Atmos. Chem. Phys., 10, 6391–6408, 2010 www.atmos-chem-phys.net/10/6391/2010/ J. T. Wilkerson et al.: Analysis of emission data from global commercial aviation 6393 airlines. The guide represents all US scheduled airlines and 2.2 Fuel burn and emissions the majority of scheduled worldwide airlines. For incomplete ETMS flights and all of ETFMS and OAG flights, trajectories After creation of the movement database described above, are generated. In 2004, trajectories were generated from sta- AEDT calculates chord-level fuel burn and emissions. This tistical distributions of cruise altitudes and horizontal tracks requires knowledge of takeoff weights, aircraft performance between origination and destination (OD) airport pairs. The data, and emission data. Validation assessments have shown distributions for a given flight-distance category (e.g., 200– that SAGE can predict fuel burn to within 3% of airline data 250 nm) were generated at Volpe by statistically analyzing a (Kim et al., 2007). The emission module then generates es- large set of ETMS flights for jet or turboprop engines, the timates of emissions based on the amount of fuel burned and percentage of each flight that occurred along the OD Great on emission indices for each species (Table 1). Values re- Circle (GC), the distance from the GC, and the probability of ported here fall within the ranges summarized by the AT- the offset distance from the GC. The resulting distributions TICA team (Lee et al., 2010). Emissions of CO2, water va- of horizontal tracks provide a “dispersion” effect around the por (H2O), and SOx (modeled as SO2) follow the fuel burn OD GC estimate. An analogous method was applied to de- since they are based strictly on total fuel composition using termine cruise altitudes. Boeing-derived EI (Hadaller and Momenthy, 1993). Emis- For 2006, the same method was used to determine cruise sions of carbon monoxide (CO), hydrocarbons (HC), and altitude, but a more accurate flight corridor matching or “air- NOx are a function of the individual segment performance ways track” method was used which better represents non- (e.g. take-off, cruise, etc) so cannot be linearly scaled by to- radar flight trajectories by learning where scheduled flights tal emissions; these species are modeled using Boeing Fuel usually fly when travelling between a given OD pair. This Flow Method 2 (BFFM2) (Baughcum, 1996b; DuBois and method determines the shortest distance for flights traveling Paynter, 2006). between common OD pairs as the horizontal track, which Changes in CO and HC EIs, are a result of improvements better reflects where these aircraft typically fly. If a valid air- made in AEDT for terminal-area fuel burn calculations and ways track did not exist between an OD pair (e.g. when track changes in EIs. The highest EI of CO is at low thrust setting, data between the two airports is non-existent), then a dis- such as during the idle (taxiing operations) or descent por- persed GC track was used as in the 2004 data. The airways tions of flight. While the terminal area represents a propor- track approach closely resembles radar data when compar- tionately small amount of the overall fuel burn, it represents ing both within AEDT, neglecting reroutes due to adverse a substantial portion of the overall CO produced. weather. This AEDT emissions inventory is similar to the European The EI for black carbon (BC) particulate matter (PM) used AERO2k database (Eyers et al., 2005). Where AEDT uses in the AEDT’s 2004 inventory was 0.2 g/kg-fuel which rep- ETMS, ETFMS if available, and OAG data as the main com- resents the EI for PM during a portion of the aircraft oper- ponents; AERO2k supplements actual flight data with sched- ation, including takeoff and climb out; however this repre- ule data from the Back Aviation database (fleet registration), sented a small portion of the activity and effectively over- and by airspace route structure information (Michot et al., predicts BC emissions. For the 2006 inventory, AEDT chose 2004). The agencies that control these databases, FAA and to use 0.035 g/kg for BC EI which is more consistent with EUROCONTROL, have increased the amount of exchange cruise emissions. Both 2004 and 2006 EI values fall be- of flight-movement information; so, both databases are ex- tween the low and high range of other inventories as sum- pected to improve in accuracy and completeness as collabo- marized by Lee et al. (2010). Recent work by Volpe and the ration continues. FAA (Wayson et al., 2009) has advanced the methodology Once flight trajectories are identified within AEDT and to estimate PM emissions by disaggregating non-volatile PM SAGE, each flight is divided into multiple linear segments, from fuel organics and sulfur-related compounds; however, or chords. Each chord includes spatial and temporal infor- this is intended for airport operations at ground level condi- mation of the chord, flight identification information, and tions rather than cruise-related operations. The methodology pertinent performance parameters such as aircraft and engine may be applicable at cruise conditions with modifications; type. The horizontal and vertical resolution of the segment however, this has not been verified. end points are 10−6 degrees and 10−4 m respectively, and the The PM and gaseous EI predicted by AEDT are continu- time is resolved to integer seconds. Each flight may consist ally improved by analysis of aircraft emissions tests which of a few dozen to several hundred chords, and the length of help characterize exhaust plumes as a function of perfor- each chord is determined by dynamic parameters such as sig- mance such as during taxi, takeoff, cruise and approach. One nificant change in aircraft speed or altitude, or in horizontal such study by The Airport Cooperative Research Program deviation. summarized emissions by engine type, thrust, and performance, atmospheric conditions, and plume age (Whitefield et al., 2008). However, like most studies of this nature, this is a ground-based experiment. There have also been several www.atmos-chem-phys.net/10/6391/2010/ Atmos. Chem. Phys., 10, 6391–6408, 2010 6394 J. T. Wilkerson et al.: Analysis of emission data from global commercial aviation

Table 1. Emissions indices. and 600 thousand segments depending on daily flight activity. Visualizing the flight emissions data enables a better un- Emission Factors 2004 2006 derstanding of spatial and temporal characteristics of the emissions, and a gridded format, rather than vector format, is H2O emissions (g/kg-fuel) 1237 1237 1 better suited for this visualization. While Volpe could have CO2-C emissions (g-C/kg-fuel) 861 862 ◦× ◦× 2 5 5 provided gridded results at 1 1 1 km resolution, gridding NO2 emissions (g/kg-fuel) 14.5 14.1 5 5 the provided vector format allowed for both the visual veri- CO emissions (g/kg-fuel) 2.57 3.61 fication of the vector data and a finer grid resolution than SO -S emissions3 (g-S/kg-fuel) 0.578 0.588 x readily available from Volpe. For this study, the individual HC (as CH ) emissions (g/kg-fuel) 0.3505 0.5205 4 flight data were converted to 0.5◦×0.5◦×0.5 km spatial grid- Organic PM emissions4 (g-POM/kg fuel) 0.015 0.015 ded data. Sulfur PM emissions3 (g S(VI)-S/kg-fuel) 0.022 0.012 Gridding the data required identifying which altitude and Black Carbon PM emissions (g-BC/kg-fuel) 0.200 0.035 horizontal bins the segment intersected and applying the ap- 1 propriate fraction of emissions to each bin. This was accom- Reported as 3155 g CO2/kg fuel for 2004, and 3159 g CO2/kg fuel plished with a three-dimensional parametric vector-plane in- in 2006. Carbon from CO obtained by molecular weight ratio: 2 tercept solution. The distance to the vector intersection with MC/MCO2 ; 2 the next longitude, latitude, or altitude plane, based on the NOx converted to NO2 in database; 3 chosen resolution, determined the fractional amount of the in 2004, 96.3% of the SOx-S was partitioned to SO -S (gas) and 2 segment emissions applied to each cell. The solution to 3.7% to S(VI)-S(particle). For 2006, 98% was partitioned to SO2- S; each gridded segment was then added to the accumulating 4 primary organic matter; global gridded dataset. Once all hourly 3-D gridded arrays 5 effective EI based on total emissions and fuel burn. Actual EI were written, the individual hourly files were interrogated for is a function of flight activity, such as take-off and cruise, so not statistics such as emissions during a season, over a particular constant. region or above a given altitude.

2.4 Regional disaggregation flight-level experiments and results from such studies are Gridding the data to the desired resolution facilitates a visual added into AEDT’s methods when appropriate. inspection of the data both spatially and temporally, and this 2.3 Vector to grid conversion global footprint illustrates where the emissions generally occur. However, understanding local emissions enables insight We analyzed 4-D inventories for 2004 and 2006, which were into the regional impacts. The global grid was parsed into re- produced by Volpe using the above methodology. The pa- gions, and the following describes the two methods used for rameters we used included the beginning and end time, lati- isolating portions of the global grid. tude, longitude, and altitude of each flight segment of every One method of aggregating emissions into world regions uses the 19 IPCC regions defined in the IMAGE model flight and the fuel use and emissions of each species during ◦ ◦ the segment. Emission data were prepared by Volpe into 366 (RIVA, 2001) which isolates regions on a 1 ×1 grid res- and 365 daily files for 2004 and 2006, respectively, where olution. They are identified in Table 2 and illustrated in flight segments, or vectors, were assigned to a specific daily Fig. 1a. The one-degree resolution was lower than that of the file corresponding to the day of the begin-time of the flight. gridded emission data, but allowed for sufficient disaggrega- Each annual dataset contains nearly a Terabyte of flight data tion of emissions into logical regions of irregular shapes that in this vector format. roughly match country or region boundaries. The segment data is appropriate for incorporating the The second method was to isolate major regions of interest emission data in a climate model that treats subgrid emission with rectangular bounding boxes. These regions were chosen plumes and contrails and their spreading due to shear and di- based on peak aviation activity or research interest, such as lution (Jacobson et al., 2010; Naiman et al., 2010); however, over the US or the North Atlantic Flight Corridor. These the large daily files are too large and cumbersome to load regions are illustrated in Fig. 1b and defined in Table 3. into computer memory. Consequently, these daily files were further divided into hourly files, which are more manageable 3 Results and of the same time-scale order as in the numerical climate study. Flight segments were broken into sub-segments and 3.1 Annual statistics stored in the appropriate hourly file in which each fraction occurred. This resulted in 8784 hourly files for 2004 and Annual totals for the fuel burned, distance travelled, and 8760 files for 2006. Each hourly file consists of between 200 computed emissions constituents are summarized in Table 4.

Atmos. Chem. Phys., 10, 6391–6408, 2010 www.atmos-chem-phys.net/10/6391/2010/ J. T. Wilkerson et al.: Analysis of emission data from global commercial aviation 6395

Table 2. IPCC regions (see Fig. 1a for graphical presentation). Table 3. Bounds for regional studies (see Fig. 1b for graphical presentation). IPCC Region Total Area % Total (103 sqkm) Box Regiona Total Area % Totalb (103 sqkm) 0 OCEAN 327 501 64.21% 1 CANADA 12 642 2.48% A United States 15 687 3.08% 2 USA 11 573 2.27% B Eastern US 7711 1.51% 3 CENTRAL AMERICA 5354 1.05% C Western US 7976 1.56% 4 SOUTH AMERICA 20 307 3.98% D Arctic Region 21 152 4.15% E Europe 6630 1.30% 5 NORTHERN AFRICA 6591 1.29% F Eastern Asia 16 126 3.16% 6 WESTERN AFRICA 12 203 2.39% G North Atlantic Flight Corridor 11 430 2.24% 7 EASTERN AFRICA 6985 1.37% H North Paciﬁc Flight Corridor 23 571 4.62% 8 SOUTHERN AFRICA 7776 1.52%

9 OECD EUROPE 6421 1.26% a Box Region Dimensions: 10 EASTERN EUROPE 1246 0.24% A) 125◦ W–66◦ W×23◦ N–50◦ N E) 12◦ W–20◦ E×35◦ N–60◦ N 11 FORMER USSR 25 157 4.93% B) 95◦ W–66◦ W×23◦ N–50◦ N F) 103◦ E–150◦ E×15◦ N–48◦ N ◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦ 12 MIDDLE EAST 7110 1.39% C) 125 W–95 W×23 N–50 N G) 70 W–5 W×40 N–63 N D) North of 66.5◦ N H) 140◦ E–120◦ W×35◦ N–65◦ N 13 SOUTH ASIA 6068 1.19% b Column does not sum to 100% 14 EAST ASIA 12 040 2.36% 15 SOUTHEAST ASIA 9576 1.88% 16 OCEANIA 13 062 2.56% when ETMS radar flight data were not available or incom- 17 JAPAN 948 0.19% plete, the 2004 inventory employed a dispersion along the 18 GREENLAND 2715 0.53% great circle to populate the horizontal tracks, while the 2006 19 ANTARCTICA 14 790 2.90% inventory employed a combination of airways track resorting Total 510 064 100.00% to a dispersed great circle only when an airways track was not available. Both datasets included about the same percentage of non-radar tracks (about 50%) which occur mostly over Eastern Europe and Asia. Both horizontal track methods at- The reported data in the table suggest fuel burn and subse- tempt to capture the true length of flights. The GC dispersion quent emissions decreased between 2004 and 2006; however method estimates flight distances on average, while the air- this is misleading, as the original 2004 data require correc- ways track method estimates the length of every flight. How- tion. Corrected 2004 data are also shown in Table 4; however, the airways method provides shorter distances overall ever, explanations and a description of corrections for this when compared to the GC dispersion method. While a global discrepancy are addressed below, respectively, after a brief comparison of the difference between methods is not avail- description of the results. Other reliable gauges of aviation able, the latter has been determined by Volpe to be more ac- activity suggest an increase between 2004 and 2006 (OAG, curate, suggesting the earlier method over-predicts distances 2007; IEA, 2009), a result similar to that found here once the and subsequent fuel burn and emissions. 2004 data are corrected. Recent studies have shown that actual flight paths can increase flight distances an average of 10% in Europe and 3.1.1 Global footprint 6–8% in the US compared with direct GC estimated distances (Kettunen et al., 2005). Both horizontal track meth- In 2006, the global aviation fleet burned 188.20 Tg (or metric ods employed by AEDT were aimed at defeating this under- tons) of fuel. Figure 2 illustrates where fuel was consumed prediction by not using a direct GC track for non-radar annually during 2004 and 2006. Several striking features flights. Kettunen further found that 70% of the underes- of the emissions distributions are evident from the figures. timation occurred within the terminal control areas (TCA) Nearly all aviation activity occurred in the Northern Hemi- which is the volume of space immediately surrounding air- sphere where 92.5% of the fuel was consumed in 2006, and ports where planes are taking off and landing. So it is possi- 69.0% of the activity occurred in the Northern mid-latitudes ble that neither the 2004 nor the 2006 inventory captured this of 30–60◦ N. The 2006 inventory was created using an up- TCA spatial uncertainty for non-radar flights. dated version of the AEDT system relative to 2004; thus, Table 4 also shows several computed global emission con- there are some visual differences between the original 2004 stituents. Some species are proportional to fuel burn, in- and the 2006 data in the local distribution of the flight tra- cluding CO2, water vapor, and SOx. Others, such as NOx, jectories; especially over Central Asia. As described earlier, sulfur, and fuel PM, are performance-based. Due to the www.atmos-chem-phys.net/10/6391/2010/ Atmos. Chem. Phys., 10, 6391–6408, 2010 6396 J. T. Wilkerson et al.: Analysis of emission data from global commercial aviation

Fig. 1. Regions of the world deﬁned by one of two methods: (a) IPCC Regions based on the IMAGE model (RIVA); (b) Bounding Box Regions identifying regions of principal activity, including the US, Europe, East Asia, Arctic, and North Atlantic and North Paciﬁc Flight Corridors.

Fig. 2. Spatial distribution of fuel burned in (a) 2004 (uncorrected) and (b) 2006. Global distributions are similar between the two years: nearly all of the emissions occurred in the Northern Hemisphere and about three out of every four grams of fuel was burned in the northern mid-latitudes. Visual differences between the two plots are largely an artifact of the different horizontal track estimation methods used between the two years. The “airways track” method used in 2006 is considered a more accurate representation of actual flight tracks. previously-mentioned database differences, a quantitative 3.1.2 Correction of 2004 data comparison of the reported global totals between uncorrected 2004 and 2006 is not appropriate. However, some differ- Other publications of aviation emission data suggest fuel ences are significant, such as with CO and BC, whose emis- burn increased between 2004 and 2006. For example: the sions changed significantly between 2004 and 2006 despite International Energy Agency (IEA) reported an increase in the corrections discussed shortly. BC decreased significantly total primary energy supply of global aviation bunkers from due to the use of a more appropriate mean-EI as described 124.2 Tg in 2004 to 135.0 Tg in 2006 with similar increases in Sect. 2.2. Future improvements of AEDT’s model are ex- in computed CO2 emissions (IEA, 2009). IEA relies on pected to include performance-based EI for BC as well to reported information, which may explain why the agency’s more accurately compute BC emissions. Changes in CO, numbers are lower than in the current study. Also the OAG and HC, are a result of improvements made in AEDT for reported continuous year-over-year growth in the number of terminal-area fuel burn calculations and changes in EIs. For seats sold of 3–5% every year from 2000–2006 (OAG, 2007). all species, the 2006 result is considered the benchmark for The 2004 fuel burn results reported here are also 8.6% EI and total results. higher than fuel burn reported for the same year in a previous SAGE inventory report (Kim et al., 2005c), which are shown in Table 5 in units consistent with this study. The

Atmos. Chem. Phys., 10, 6391–6408, 2010 www.atmos-chem-phys.net/10/6391/2010/ J. T. Wilkerson et al.: Analysis of emission data from global commercial aviation 6397

Table 4. Total annual emissions from global commercial aviation, 2004 and 2006.

2004 2006 Reported Correctede (Comments) Distance traveled (Billion km) 41.42 38.68 Total number of ﬂights (Million) 33.13 31.26 Fuel burned (Tg) 205.68 174.06 188.20 In Northern Hemisphere 92.7% 92.5% (% of Fuel burned) In Northern Mid-Latitudes 68.9% 69.0% (30◦ N–60◦ N)

H2O emissions (Tg) 254.42 215.31 232.80 a CO2-C emissions (Tg) 177.09 150.06 162.25 (C from CO2) b NO2 emissions (Tg) 2.987 2.456 2.656 (NOx as NO2) CO emissions (Tg) 0.529 0.628 0.679 c SOx-S emissions (Tg) 0.119 0.102 0.111 (S from SOx) HC (as CH4) emissions (Tg) 0.072 0.090 0.098 Organic PM emissions (Tg) 0.1516 0.0026 0.0030d Sulfur PM emissions (Tg) 0.0046 0.0021 0.0023 Black Carbon PM emissions (Tg) 0.0386 0.0061 0.0068 a Carbon from CO2 obtained by molecular weight ratio: MC/MCO2 ; b NOx Converted to NO2 in database; c sulfur from SOx obtained by molecular weight ratio: MS/MSO2 ; d original 2006 inventory included 0.135 Tg organic PM, modified here to be consistent with BC emissions; e corrected global fuel burn results for 2004 without over-count but with 2006-equivalent trackways generation. Corrected emissions calculations rely strictly new preliminary fuel burn and 2006 EI values from Table 1. difference between the two 2004 reports is in small part due Table 5. Annual emissions reported in previous SAGE study. to better representation of unscheduled flights in the current report. However, it is largely due to an over-count which is Previous Study Resultsa best explained in Sect. 3.2 below. With the recent discovery 2000 2001 2002 2003 2004 of this significant over-count (only in the 2004 inventory), Fuel burned (Tg) 181 170 171 176 188 Volpe has reprocessed the 2004 emissions. AEDT’s capabil- b CO2-C emissions (Tg) 156.11 146.28 147.10 152.01 162.11 ities are continually updated, so the new 2004 inventory has c NO2 emissions (Tg) 2.510 2.350 2.410 2.490 2.690 been processed very much like the 2006 inventory by using the airways track method for non-radar flights. As such, the a Data from (Kim, 2005); corrected 2004 global results in Table 4 are directly compara- b carbon from CO2 obtained by molecular weight ratio: MC/MCO2 ; c ble with the 2006 global results for trends in aviation activity. NOx Converted to NO2 in database. Corrected results provided by Volpe show the total global fuel burn for 2004 was 174.06 Tg, compared with 188.20 Tg in 2006, which indicates that fuel burn increased from 2004 3.2 Temporal emissions to 2006 as expected. This suggests an annual growth rate of 3.95%, which agrees well with OAG and IEA data sources. Figure 4a shows the daily annual carbon from CO2 emissions The corrected speciated emissions for 2004 are estimated us- (CO2-C) for 2004 and 2006. Since most emissions occurred ing the new fuel burn total and the 2006 EI values from Ta- in the Northern Hemisphere, the annual temporal distribution ble 1. The 2006 and the corrected 2004 fuel burn totals match was dominated by the Northern Hemisphere seasons. Air well with other recent inventory studies such as AERO2K, travel increased in early April and continued through Octo- TRADEOFF, and FAST (Eyers et al., 2005; Lee et al., 2005; ber and dropped off during the winter. The increase in air Gauss et al., 2006) as shown in Fig. 3. These and older stud- traffic in summer led to a peak in daily emissions from July ies have been summarized well by the ATTICA team (Lee through August. Mid-winter activity dropped to a low in Jan- et al., 2010). Since previous horizontal track methodologies uary and February. However, the 2004 data set contains an over-predicted flight lengths, the 2006 and corrected 2004 artificial over-count of operations all year and a daylight sav- fuel burn results represent the best estimates of global avia- ings coding error is evident from April through October. tion consumption for these years. www.atmos-chem-phys.net/10/6391/2010/ Atmos. Chem. Phys., 10, 6391–6408, 2010 6398 J. T. Wilkerson et al.: Analysis of emission data from global commercial aviation

Fig. 3. Fuel burn estimates by year from recent studies: TRADE- OFF, FAST, AERO2K, SAGE 1.5 compared with current AEDT results.

For flight schedule information, radar-based (ETMS) and schedule based (OAG) were pulled together globally. The attempt to merge these data sources was not perfect and resulted in some double-counts throughout the year, primarily over North America where radar coverage occurred. This over-count was also present in the previous SAGE emissions report referred to above. While most of the difference between the current corrected 2004 study and the older 2004 SAGE study represents the difference in track methodologies, a smaller undetermined amount is a combination of this underlying double-count and the continual improvements to the database. During daylight savings period, a bug in the merging resulted in a much more significant double-count during these months. We can estimate the daylight savings error by es- timating the effect of the jump in emissions in each of the seven months. From Fig. 4a, the uncorrected 2004 CO2- C daily mean emissions are about 0.01 Tg per day higher than computed for 2006 during non-daylight savings months. Taking this into account when comparing the two curves: un- Fig. 4. Annual 2004 (uncorrected) and 2006 CO2-C emissions. (a) corrected 2004 data are about 14% higher than expected in total daily emissions, (b) total emissions per day of the week, (c) April and May, 13% higher in June, 8% higher in July and average contributions for each hour of the day. August, and 11% higher in September and October. Com- bined, the daylight savings portion of the over-count is estimated at about 6.2% or 11 Tg of CO2-C for the year. Sub- Again, the data here should not be used to indicate emissions tracting this daylight savings error reduces the reported 2004 trends between 2004 and 2006, but the results here are use- CO2-C emissions to about 166 Tg globally. After subtracting ful to researchers already using either dataset for global or the daylight saving error, there still remains a smaller over- regional climate studies. count of undetermined amount. The remaining difference The daily emission curves in Fig. 4a show an oscillation between the two reports is a combination of the inclusion of which illustrates the weekly periodicity. Integrating all emis- unscheduled flights and other regular improvements in the sions by day of the week shows aviation activity peaked process. on Thursday and Friday and dropped off for the weekend The known over-counts were resolved before the 2006 (Fig. 4b). The daily and weekly profiles were tallied relative data were computed, by using ETMS exclusively in regions to Universal Time Coordinate (UTC), which led to an inves- where radar coverage occurred (primarily North America). tigation of emission contributions by the hour. A compila- Thus, the 2006 data set more accurately captures seasonal tion of the hourly global data resulted in the 24-h profile in trends, with peak operation count in the Northern Hemi- Fig. 4c, which shows annual average emissions per hour. The sphere summer months and lowest during the winter months. over-counts affected primarily North America were radar

Atmos. Chem. Phys., 10, 6391–6408, 2010 www.atmos-chem-phys.net/10/6391/2010/ J. T. Wilkerson et al.: Analysis of emission data from global commercial aviation 6399

Fig. 5. Composite of hourly column CO2-C emissions through the daily cycle on 25 August, 2006. Time is measured relative to UTC. Activity at 00:00-UTC shows evening traffic in Asia. As time progresses, morning rush hours in Europe, then the United States are evident. Toward the end of the day in the US, activity moves over the Pacific toward Asia and redeye flights back to Europe to start the cycle over again. coverage existed, which is evident by the behavior of the The composite image in Fig. 5 shows the integrated an- 2004 hourly emissions curve in Fig. 4c. As shown earlier, nual emissions incrementally through the day, again relative less fuel was, in fact, burned in 2004 than in 2006. It follows to UTC. Hour 00:00 UTC is midnight in London, and after- that comparisons between the two years should also show noon in Asia where the industry is still very active. Over less per day or per hour than in 2006. That said, as activ- the next several hours, planes from Asia begin making their ity rapidly increased in North America, beginning around way toward Europe and arrive in time for the local morning 12:00 UTC, the 2004 curve increases faster than the 2006 rush hour. By hour 08:00 UTC, rush hour in Great Britain curve; thus, indicating the over count is embodied primar- has ensued and aviation activity over Europe is increasing ily in North American activity. This evidence does not show quickly. 12:00 UTC is 07:00 a.m. Eastern Standard Time up in Fig. 4b, since the data were aggregated by day. along the US Eastern Seaboard; and aircraft racing across the North Atlantic join an increasing amount of aircraft orig- inating in Eastern US. Over the next three hours, the rest of North America begins to add to aircraft activity as Europe begins to slow down at the end of the work day. Toward the www.atmos-chem-phys.net/10/6391/2010/ Atmos. Chem. Phys., 10, 6391–6408, 2010 6400 J. T. Wilkerson et al.: Analysis of emission data from global commercial aviation

Table 6. Flight length and duration for 2006 short-haul and long-haul ﬂights.

Global Statistics for 2006 Total Percent of Total Total (100%) Distance traveled (Billion km) 38.68 100.00% Total number of flights (Million) 31.26 100.00% a CO2-C emissions (Tg) 162.25 100.00% Average flight length (km) 1237.19 Average flight time (hr) 2.06 CO2-C emissions (kg/km) 4.20 Short-Haul (Less than 3 h flight time) Distance traveled (Billion km) 20.85 53.92% Total number of flights (Million) 26.62 85.17% CO2-C emissions (Tg) 64.34 39.65% Average flight length (km) 783.28 Average flight time (hr) 1.50 CO2-C emissions (kg/km) 3.09 Medium-Haul (Between 3 and 6 h flight time) Distance traveled (Billion km) 9.61 24.85% Total number of flights (Million) 3.48 11.15% CO2-C emissions (Tg) 36.74 22.65% Average flight length (km) 2759.01 Average flight time (hr) 3.96 CO2-C emissions (kg/km) 3.82 Long-Haul (More than 6 h flight time) Distance traveled (Billion km) 8.21 21.23% Total number of flights (Million) 1.15 3.69% CO2-C emissions (Tg) 61.17 37.70% Average flight length (km) 7117.94 Average flight time (hr) 9.28 CO2-C emissions (kg/km) 7.45 a Carbon from CO2 obtained by molecular weight ratio: MC/MCO2 . end of the day in the US, aircraft along the east coast begin red-eye flights toward Europe, and aircraft along the Pacific coast head out over the Pacific Ocean toward Asia where the activity starts all over again the next day.

3.3 Global CO2-C emissions

The 2006 dataset describes 31.26 million flights travelling a total of 38.68 billion km, emitting about 162.25 Tg of carbon from CO2 (CO2-C). These emissions represent about 2– 3% of global anthropogenic CO2 emissions (Sausen et al., 2005). In 2006, the total mean flight was 1237.19 km and 2.06 h, which is only slightly longer than a typical flight between San Francisco and Los Angeles, California. This indicates the dataset is dominated by short-haul flights which, by one definition, are those that are less than three hours in Fig. 6. Percent distribution of all flights for 2006. There are some duration. Long-haul flights are those lasting more than six suspect flights in the data that cover 6–10 thousand km in only a hours, and flights lasting between three and six hours are few hours, as well as those that take many hours to cover relatively considered medium-haul flights. Aggregating the data by short distances; however these represent an extremely small fraction these rules provides the results in Table 6. Short-haul flights of the data. represent 85.17% of the total number of flights. This is il-

Atmos. Chem. Phys., 10, 6391–6408, 2010 www.atmos-chem-phys.net/10/6391/2010/ J. T. Wilkerson et al.: Analysis of emission data from global commercial aviation 6401

Fig. 7. Spatial distribution of carbon emissions in CO2. (a) Global column total; (b) Latitude profile; (c) Longitude profile. lustrated in the length-duration plot in Fig. 6, which shows All three of these primary regions, and the dominant flight the percentage of flights for a given distance and duration corridors between them, occur in the Northern Hemisphere bin. Short-haul flights account for about half (53.92%) of and almost entirely within northern mid latitudes. Weather the total annual distance travelled by all commercial aviation circulation patterns in the lower latitudes do not often ex- and emit 39.65% of the total CO2-C. By contrast, long-haul change between the Northern and Southern Hemispheres, flights emit about the same total CO2-C (37.70%); yet they and winds in the mid latitudes are typically strong and spread account for only 3.69% of the total number of flights and emissions in this region quickly (Zhao and Li, 2006). Thus, 21.23% of total distance travelled annually. Long-haul air- since these three regions enjoy the predominant share of avi- craft must carry the fuel for a longer mission; this weight ation activity, they also suffer from the majority of global leads to more fuel burn and twice the rate of emissions of an aviation emissions and subsequent effects. average short-haul flight. The provided AEDT flight data are with respect to mean Figure 7 illustrates annual 2006 carbon emissions from sea level (MSL), assuming an international standard atmo- CO2, which was 2–3% of total anthropogenic CO2-C emis- sphere pressure profile. Emissions at altitudes above 7 km sions. Not surprisingly, it is very much like the fuel burn can be considered cruise-related emissions, while those be- footprint for the same year. However, the data are presented low 7 km include shorter flights and local emissions associ- here as emissions per unit area to better understand some of ated with airport activity such as taxi, take-off, and landing. the regional effects described in Sect. 3.4 below. Latitude and 74.6% of all aviation fuel was burned above 7 km, where Longitude zonal profiles are included with the figure. The emissions have a longer residence time than those emitted latitude plot (Fig. 7b) shows the disparity between Northern near the surface. The global vertical profile of 2006 CO2- and Southern Hemisphere emissions, with a peak of nearly C emissions (Fig. 8a) shows some activity near zero MSL, 2 ◦ 2400 kg/km CO2-C occurring around 40 N. The longitude which is associated with airport activity and take-off/landing. plot (Fig. 7c) highlights three regions of significant activity There is a gradual increase in emissions up through 7 km, that account for more than half of the global 2006 CO2-C then a dramatic increase above 9 km. Peak emissions oc- emissions: the United States, Europe, and East Asia. curred between 10–12 km in the upper troposphere and lower stratosphere (UTLS) where contrails predominantly form. www.atmos-chem-phys.net/10/6391/2010/ Atmos. Chem. Phys., 10, 6391–6408, 2010 6402 J. T. Wilkerson et al.: Analysis of emission data from global commercial aviation

Table 7. IPCC Regional CO2-C Emissions for 2004 (uncorrected) and 2006.

2004 (uncorrected) 2006 a 2 a 2 IPCC Region CO2-C (Tg) % Total kg/km Above 7km CO2-C (Tg) % Total kg/km Above 7 km 0 OCEAN 35.05 19.79% 107 98.7% 29.99 18.48% 92 99.2% 1 CANADA 9.56 5.40% 756 87.6% 6.80 4.19% 538 87.5% 2 USA 50.72 28.64% 4,383 66.6% 41.77 25.74% 3609 65.1% 3 CENTRAL AMERICA 4.43 2.50% 827 66.7% 3.26 2.01% 610 68.6% 4 SOUTH AMERICA 5.04 2.85% 248 72.2% 4.28 2.64% 211 68.2% 5 NORTHERN AFRICA 1.42 0.80% 216 84.2% 1.57 0.97% 239 75.7% 6 WESTERN AFRICA 1.23 0.69% 101 83.4% 1.21 0.74% 99 79.7% 7 EASTERN AFRICA 0.75 0.43% 108 82.5% 0.76 0.47% 109 80.6% 8 SOUTHERN AFRICA 1.01 0.57% 130 74.2% 1.06 0.65% 137 71.6% 9 OECD EUROPE 22.52 12.71% 3507 63.2% 23.95 14.76% 3730 60.0% 10 EASTERN EUROPE 2.67 1.51% 2,146 89.8% 3.55 2.19% 2847 86.0% 11 FORMER USSR 10.22 5.77% 406 91.2% 8.12 5.00% 323 92.3% 12 MIDDLE EAST 5.35 3.02% 752 77.1% 6.10 3.76% 858 71.6% 13 SOUTH ASIA 3.55 2.01% 586 81.5% 4.36 2.69% 719 78.8% 14 EAST ASIA 10.45 5.90% 868 74.4% 11.70 7.21% 971 68.3% 15 SOUTHEAST ASIA 5.50 3.11% 575 69.4% 5.76 3.55% 601 63.9% 16 OCEANIA 3.26 1.84% 249 71.3% 3.31 2.04% 254 68.9% 17 JAPAN 3.86 2.18% 4071 55.6% 4.45 2.74% 4694 56.8% 18 GREENLAND 0.49 0.28% 181 99.3% 0.25 0.15% 90 99.5% 19 ANTARCTICA 0.00092 0.0005% 0.0625 100.0% 0.00061 0.0004% 0.0415 100.0% Global Totals 177.09 100.00% 347.20 77.1% 162.25 100.00% 318.10 74.6% a Carbon from CO2 obtained by molecular weight ratio: MC/MCO2 .

Fig. 8. Altitude proﬁle of and altitude-latitude zonal plot of 2006 annual CO2-C emissions.

An interesting double peak occurs in the UTLS, which can 3.4 Regional CO2-C emissions be attributed to East-West and North-South ﬂight clearance altitudes for safe passing. The altitude zonal plot (Fig. 8b) Table 7 lists the CO2-C emissions for each of the IPCC SRES shows the bulk of emissions occurred over the northern mid regions, as well as the percent of the regional emissions that latitudes; emissions in this region predominantly occur be- occurred above 7 km. These are emissions occurring within low the mean tropopause where they will contribute to tro- the boundaries of each region without regard for origination pospheric weather patterns and cloud formation. or destination of the air trafﬁc. Despite covering only 2.27% of the total surface area, 25.74% of the emissions occurred within the US borders (including Alaska) in 2006.

Atmos. Chem. Phys., 10, 6391–6408, 2010 www.atmos-chem-phys.net/10/6391/2010/ J. T. Wilkerson et al.: Analysis of emission data from global commercial aviation 6403

Table 8. Box Region CO2-C Emissions.

2004 (Uncorrected) 2006 a b 2 a b 2 Box Region CO2-C (Tg) % Total kg/km Above 7km CO2-C (Tg) % Total kg/km Above 7 km A United States 55.68 31.44% 3549 68.7% 45.18 25.51% 2880 67.2% B Eastern US 33.98 19.19% 4407 65.8% 27.43 15.49% 3558 64.2% C Western US 21.70 12.25% 2720 73.3% 17.75 10.02% 2225 71.7% D Arctic Region 2.44 1.38% 115.6 97.2% 1.14 0.64% 53.9 95.2% E Europe 24.75 13.97% 3733 67.9% 25.82 14.58% 3895 64.4% F Eastern Asia 17.79 10.05% 1103 75.1% 19.57 11.05% 1214 71.3% G NAFC 13.94 7.87% 1220 96.8% 11.46 6.47% 1003 96.9% H NPFC 12.85 7.25% 545 49.9% 11.15 6.30% 473 47.5% Global Totals 177.09 100.00% 347.20 77.1% 162.25 100% 318.10 74.6% a Carbon from CO2 obtained by molecular weight ratio: MC/MCO2 ; b column does not sum to 100%.

Differences in track methodologies employed by AEDT dominance: the Eastern US region receives about two thirds between 2004 and 2006 inventories will not allow a quantita- of the US emissions with a concentration of more than 11 tive comparison of spatial emissions; however, these differ- times the 2006 global average emissions concentration. The ences can drive significant regional variations. For example, Highest concentration of emissions is over Europe, which re- in Eastern Europe, Table 7 shows emissions increased sig- ceives more than 12 times the global average concentration. nificantly, whereas emissions in neighboring former USSR There is no evident east-west inflection in the European pro- exhibited similar numerical decrease. The 2006 inventory file, so this region was not sub-divided; however, the northern included data from ETFMS, which includes Eastern Europe; part of the region does see slightly more activity. thus, greatly expanding radar-provided schedule and flight The East Asia box region encompasses about the same data. By including ETFMS, which enabled the use of more area as the US box, but received less than half of emissions airways tracks instead of GC dispersion estimates, East- when compared with the US box. The activity was predom- ern Europe was more comprehensively represented in 2006, inantly along a Northeast-Southwest orientation. China’s re- compared with 2004. Again, the methodology used for the ported aviation activity grew at an average of 14.5% per year 2006 data set is considered a more accurate representation of through 2007 (CAAC, 2009). Whether this is entirely growth actual flight trajectories. The over-count in the 2004 data set or partially from increased accountability, aviation activity predominately affects ETMS radar flights over North Amer- and subsequent emissions within this region are expected to ica, so such comparisons of regional emissions, especially continue growing over the next few years. over Europe, are appropriate. The global footprint also shows two dominant flight cor- The Box regions are slightly larger than similar IPCC re- ridors: North Atlantic and North Pacific Flight Corridors gions, which allow for less dependence of results on horizon- (NAFC, NPFC). The NAFC box captures the dominant traf- tal track estimation methods. The purpose of these regions is fic exchange between Europe and the US. The NPFC box to highlight areas of significant aviation activity. From the covers a larger area so the analysis is less effective when iso- global longitudinal profile in Fig. 7c, it is apparent that the lated within a single rectangular box. Despite this expansive continental US receives more emissions than any other re- coverage, there are still other regular Pacific flight corridors gion (25.5%), followed closely by Europe (14.6%) and then which were not captured within the box, such as between Eastern Asia (11.1%). These three regions account for only Japan and Hawaii. Comparing the NAFC and NPFC, more 7.5% of global surface area, but receive over half, 51.1%, of total kilometers were flown within the NPFC bounds, but all aviation CO2-C emissions. These and other regions are the NAFC received nearly two times higher concentration of summarized in Table 8. CO2-C emissions. Also, only a third of the NPFC emissions Unlike the SRES US region, the US box region is essen- occurred above 7 km, while nearly all of the NAFC emis- tially the continental US but includes activity surrounding its sions occurred above 7 km. These and other totals are shown borders. This region received about one quarter of all global in Table 8. emissions in 2006. Emissions within the region were dom- Another region of interest is the Arctic. The total emis- inant in the eastern half of the country, which has more air- sions over this region during 2006 were about 0.64% of the ports and a larger population than the western half. Splitting global total, resulting in a mean emission concentration about the region evenly into east and west illustrates the eastern 1/6 per unit area as the global mean. Yet, in 2004, the www.atmos-chem-phys.net/10/6391/2010/ Atmos. Chem. Phys., 10, 6391–6408, 2010 6404 J. T. Wilkerson et al.: Analysis of emission data from global commercial aviation

Table 9. Speciation proﬁle for emitted hydrocarbons (FAA, 2008).

Species Mass Fraction Species Mass Fraction Species Mass Fraction Ethylene 0.15459 1-Nonene 0.00246 Methanol 0.01805 Acetylene 0.03939 n-Nonane 0.00062 Formaldehyde (FAD) 0.12308 Ethane 0.00521 Isopropylbenzene 0.00004 Acetaldehyde (AAD) 0.04272 Propylene 0.04534 n-Propylbenzene 0.00067 Acetone 0.00369 Propane 0.00078 m-Ethyltoluene 0.00193 Propionaldehyde 0.00727 Isobutene/1-Butene 0.01754 p-Ethyltoluene 0.00080 Crotonaldehyde 0.01291 1,3-Butadiene 0.01687 1,3,5-Trimethylbenzene 0.00068 Butyraldehyde 0.00148 cis-2-Butene 0.00210 o-Ethyltoluene 0.00082 Benzaldehyde 0.00470 3-Methyl-1-butene 0.00140 1,2,4-Trimethylbenzene 0.00438 Isovaleraldehyde 0.00041 1-Pentene 0.00776 1-Decene 0.00185 Valeraldehyde 0.00306 2-Methyl-1-butene 0.00174 n-Decane 0.00320 o-Tolualdehyde 0.00287 n-Pentane 0.00198 1,2,3-Trimethylbenzene 0.00133 m-Tolualdehyde 0.00347 trans-2-Pentene 0.00359 n-Undecane 0.00444 p-Tolualdehyde 0.00060 cis-2-Pentene 0.00276 n-Dodecane 0.00462 Methacrolein 0.00536 2-Methyl-2-butene 0.00185 n-Tridecane 0.00535 Glyoxal 0.01816 4-Methyl-1-pentene 0.00086 C14-alkane 0.00186 Methylglyoxal 0.01503 2-Methylpentane 0.00408 C15-alkane 0.00177 Acrolein 0.02449 2-Methyl-1-pentene 0.00043 n-tetradecane 0.00416 C-10 parafﬁns 0.14157 1-Hexene 0.00736 C16-alkane 0.00146 C-10 olefﬁns 0.05663 trans-2-Hexene 0.00037 n-pentadecane 0.00173 Decanal 0.05663 Benzene 0.01681 n-hexadecane 0.00049 Dodecenal 0.02831 1-Heptene 0.00438 C18-alkane 0.00002 n-Heptane 0.00064 n-heptadecane 0.00009 Toluene 0.00642 Phenol 0.00726 1-Octene 0.00276 naphthalene 0.00541 n-Octane 0.00062 2-methyl naphthalene 0.00206 Ethylbenzene 0.00174 1-methyl naphthalene 0.00247 m-Xylene/p-Xylene 0.00282 dimethylnapthalenes 0.00090 Styrene 0.00309 C4-Benzene+C3-aroald 0.00656 o-Xylene 0.00166 C5-Benzene+C4-aroald 0.00324

Note: Conversion factor from THC to TOG (THC is in methane equivalent) Turbine 1.1571 fractional emissions and the emissions per unit area over the Emissions over the Arctic were distributed uniformly hor- Arctic were twice those of 2006 (Table 8). This difference izontally throughout the polar region (Fig. 9), and over 95% between the two years is likely due primarily to the differ- were emitted above 7 km, where they have a longer residence ence between horizontal track generation methods. Flights time (due to greater stability in the upper troposphere and between Europe and North America through the NAFC pass lower stratosphere, where Polar emissions occur) than those close to the Arctic Circle. If a horizontal airways track does emitted near the surface at other latitudes. For comparison, not cross into the polar region, then none of the emissions in 2006 only 67.2% of emissions over the US and 64.4% from any ﬂight along that trajectory can contribute to the ge- of emissions over Europe occurred above 7 km. Circulation ographical region totals. However, the dispersion method patterns in the upper mid-latitudes trap polar emissions and used in the 2004 data allows some of the ﬂights between a push mid-latitude emissions further north (such as those from given OD pair to pass further north and contribute to the re- NAFC) into the polar region (Forster et al., 2003). Any emis- gion totals. There are several near-Arctic corridors that can sions deposited within or near the Arctic Circle are likely to contribute directly to polar emissions depending on the GC accumulate over the polar region, which further increases the method. polar emission concentrations and potential climate impacts

Atmos. Chem. Phys., 10, 6391–6408, 2010 www.atmos-chem-phys.net/10/6391/2010/ J. T. Wilkerson et al.: Analysis of emission data from global commercial aviation 6405

Fig. 9. Arctic aircraft CO2-C emissions per unit area for (a) 2004 (uncorrected) and (b) 2006. on this very sensitive region. So despite different track generation methods, climate effects are likely to be similar when comparing impacts to the Arctic region.

3.5 Other species Fig. 10. Global 2006 column-integrated emissions of other species:(a) NOx-as-NO2, (b) SOx, (c) Black carbon. In general, other emitted species have the same over-all footprint as CO2-C; however, for convenience, NOx-as-NO2, 2006 inventory is signiﬁcantly less than that used in 2004. SOx-S, and BC emissions are shown in Fig. 10. Within Older inventories have used an EI equivalent to take-off and AEDT and SAGE, NOx emissions are based on the ICAO climb-out (0.20 g-BC/kg-fuel). Since 74.6% of fuel is burned Engine Emissions Databank, which reports on an NO2 mass above 7 km, this BC EI most likely signiﬁcantly over predicts basis. So emissions of NO, NO2, and HONO are reported total BC. The 2006 inventory used an EI more appropriate for as emissions of NO2 mass-equivalence (Kim et al., 2005b). cruise performance (0.035 g-BC/kg-fuel) which better repre- Similarly, SOx is reported as SO2 in the database, and sulfur sents BC emissions; however, it probably under-predicts total from SO2 is obtained by the molecular weight ratio of S to BC emissions. SO2 (0.50). Both NO2 and SOx-S are proportional to fuel burn. However, BC emissions are also a function of engine dynamics. As described in Sect. 2.2, The BC EI used for the www.atmos-chem-phys.net/10/6391/2010/ Atmos. Chem. Phys., 10, 6391–6408, 2010 6406 J. T. Wilkerson et al.: Analysis of emission data from global commercial aviation

Unburned fuel or hydrocarbon emissions can be assumed speed rails may offer a means for offsetting a subset of these to be speciated according to the turbine-engine speciation flights and the associated emissions. profile given in Table 9 (Knighton et al., 2009). With this We have also shown the temporal distribution of emissions profile, the bulk of hydrocarbon emissions are in the form of on a weekly, daily and hourly basis. The seasonal peak oc- ethylene, formaldehyde, acetaldehyde, acetylene, propene, curred in July and August, with a decrease from Novem- C-10 paraffins, C-10 olefins, decanal, dodecanal, benzene, ber through March. During an average week, peak activ- butadiene, and butene, among others. NOx emissions specia- ity occurred on Thursday and Friday and was slowest on tion to NO-NO2-HONO is assumed to be a function of thrust Monday and Tuesday. The hourly distribution of emissions and engine type (Wood et al., 2008). was lowest at about hour 07:00 UTC and quickly ramped up through about 15:00 UTC accounting for Western European and North American rush hours. The hourly emissions re- 4 Summary mained high until the end of the Asian evening rush around 02:00–04:00 UTC. Climate impacts from CO -C, other greenhouse gases, and We have analyzed global commercial aviation emissions 2 particles, including from aviation have been studied signifi- from 2004 and 2006 in total and disaggregated by regions. In cantly to date. However, the potential impacts of aviation on 2006, the global commercial aircraft fleet flew 31.26 million climate are unique since most of the emissions occur at al- flights, burned 188.20 million metric tons of fuel and covered titudes where other anthropogenic sources are absent. The 38.68 billion kilometers. We have also provided corrected effects of aviation on stratospheric ozone and global climate 2004 total fuel burn of 174.06 million metric tons, which in- from persistent contrails and contrail-induced cirrus clouds dicates an annual growth in fuel burn of 3.95% from 2004 could be significant, but there are large uncertainties in relat- to 2006. This growth rate compares well with other aviation ing aviation emissions to changes in radiative forcing or sur- inventories and related activity. We have also estimated the face temperature from contrail-associated pathways. Know- effect of an over-count error due to a daylight savings bug ing where the emissions occur is the first step in computing in uncorrected 2004 spatial data. Remaining differences be- the potential impacts. Data presented here support a continu- tween the uncorrected 2004 and 2006 results are due to meth- ing effort to quantify the effects of aircraft exhaust on climate ods in horizontal track generation, other methodologies and and global air pollution. database improvements, and some residual double counting. Over-count errors occurred where there was radar coverage, Acknowledgements. This work was supported by the Partnership primarily in North America and Western Europe. for AiR Transportation Noise & Emissions Reduction (PARTNER) We have shown that different horizontal track methods can and the Federal Aviation Administration (FAA) under award impact the quantification of regional emissions and trends. number DTFAWA-05-D=0006. Any opinions, findings, and When radar data are not available, AEDT employed a dis- conclusions or recommendations expressed in this material are persed Great Circle method in 2004 and an airways track those of the authors and do not necessarily reflect the views of method in 2006, with the latter being more representative of PARTNER or the FAA. actual flight activity. The commercial aviation fleet emitted a total of 162.25 Tg Edited by: M. C. Facchini of CO2-C throughout 2006. The US, Europe, and Asia were subjected to 51.1% of these emissions despite covering only References 7.5% of the global surface area. The global average for CO2- C emissions per unit area was 318.1 kg/km2 in 2006. The Baughcum, S. L., Henderson, S. C., and Tritz, T. G.: Sched- Arctic received only 0.6% of the total emissions, but the per uled Civil Aircraft Emissions Inventories for 1976 and 1984: unit area emissions were about one sixth of the 2006 global Database Development and Analysis, NASA CR 4722, 1996a. average. Typical wind patterns in the upper mid-latitudes Baughcum, S. L., Tritz, T. G., Henderson, S. C., and Pickett, D. tend to trap these emissions over the arctic and push mid- C.: Scheduled Civil Aircraft Emissions Inventories for 1992: latitude emissions into the arctic risking significant conse- Database Development and Analysis, NASA CR 4700, 1996b. quences to this area of high sensitivity to climate change. CACC: General Administration of Civil Aviation in China, In 2006, an average flight covered 1237.2 km in 2.06 hours China’s Airlines to Carry 270 mln Passengers in 2010, avail- and produced 4.2 kg/km of CO -C, which indicates a domi- able online at: http://www.caac.gov.cn/English/News/200708/ 2 t20070831 7717.html, last access: 19 August 2009. nance of short-haul flights in the annual datasets. Short-haul DuBois, D. and Paynter, G. C.: Fuel Flow Method2 for Estimat- flights represented 85.2% of all commercial flights and ac- ing Aircraft Emissions, SAE Technical Paper Series, Washing- counted for 53.9% of the total distance travelled in 2006. ton, DC, 2006-01-198, 2006. These flights indicate a potential for transportation platform Eyers, C. J., Addleton, D., Atkinson, K., Broomhead, M. J., Chris- switching onto trains or buses. With improved pricing, poli- tou, R. A., Elliff, T. E., Falk, R., Gee, I. L., Lee, D. S., Marizy, cies, or incentives, existing transit systems or future high- C., Michot, S., Middel, J., Newton, P., Norman, P., Plohr, M.,

Atmos. Chem. Phys., 10, 6391–6408, 2010 www.atmos-chem-phys.net/10/6391/2010/ J. T. Wilkerson et al.: Analysis of emission data from global commercial aviation 6407

Raper, D. W., and Stanciou, N.: AERO2k Global Aviation Emis- Lee, D. S., Owen, B., Graham, A., Fichter, C., Lim, L. L., and Dim- sions Inventories for 2002 and 2025, QinetiQ Ltd, Farnborough, itriu, D.: Allocation of International Aviation Emissions from HampshireQINETIQ/04/01113, 2005. Scheduled Air Traffic – Present Day and Historical (Report 2 of Forster, C., Stohl, A., and James, P.: The residence times of aircraft 3), Manchester Metropolitan University, Centre for Air Transport emissions in the stratosphere using a mean emission inventory and the Environment, Manchester, UK., 2005. and emissions along actual flight tracks, J. Geophys. Res., 108, Lee, D. S., Pitari, G., Grewe, V., Gierens, K., Penner, J. E., Petzold, 8524–8535, 2003. A., Prather, M. J., Schumann, U., Bais, A., Berntsen, T., Iachetti, Gauss, M., Isaksen, I. S. A., Lee, D. S., and Søvde, O. A.: D., Lim, L. L., and Sausen, R.: Transport impacts on atmosphere Impact of aircraft NOx emissions on the atmosphere – trade- and climate: Aviation, Atmospheric Environment, Atmos. Envi- offs to reduce the impact, Atmos. Chem. Phys., 6, 1529–1548, ron., doi:10.1016/j.atmosenv.2009.06.005, in press, 2010. doi:10.5194/acp-6-1529-2006, 2006. Lee, J. J., Waitz, I. A., Kim, B. Y., Fleming, G. G., Maurice, L. Hadaller, O. J. and Momenthy, A. M.: Characteristics of Future Q., and Holsclaw, C. A.: System for assessing Aviation’s Global Aviation Fuels, in: Transportation and Global Climate Change, Emissions (SAGE), Part 2: Uncertainty assessment, Transport. edited by: Greene, D. L. and Santini, D. J., American Council Res.; Part D: Transportation and Environment, 12D, 381–395, for an Energy-Efficient Economy, Washington, DC, 1993. 2007. ICAO, International Civil Aviation Organization, Forecasts of Malwitz, A., Kim, B. Y., Fleming, G. G., Lee, J. J., Balasubra- Scheduled Passenger Traffic: Long-Term Traffic and Aircraft manian, S., Waitz, I. A., Klima, K., Locke, M., Holsclaw, C. Movement Forecasts, available online at: http://www.icao.int/ A., Morales, A., McQueen, E., and Gillette, W.: System for as- icao/en/atb/ead/fep/Longterm.htm, 2007. sessing Aviation’s Global Emissions (SAGE) Version 1.5: As- IEA: CO2 Emissions from Fuel Combustion, Highlights (2009 Edi- sessment Validation Assessment, Model Assumptions and Un- tion), International Energy Agency (IEA), 2009. certainties, FAA, Ed., Office of Environment and Energy, 36., Jacobson, M. Z., Wilkerson, J. T., Naiman, A. D., and Lele, S. 2005. K.: The effects of aircraft on climate and pollution – Part 1: Michot, S., Elliff, T., Fleming, G. G., Kim, B. Y., Holsclaw, C. A model that treats the subgrid evolution of discrete size- A., Locke, M., and Morales, A.: Flight Movement Inventory: and composition-resolved contrails from all commercial flights SAGE-AERO2K, Air Traffic Control Quarterly, 12, 125–145, worldwide, Atmos. Chem. Phys. Discuss., submitted, 2010. 2004. Kettunen, T., Hustache, J.-C., Fuller, I., Dan, H., Bonn, J., and MOCA, India Ministry of Civil Aviation; 2006–07 An- Knorr, D.: Flight Efficiency Studies in Europe and the United nual Report: http://civilaviation.nic.in/reports/an rep0607eng/ States, 6th USA/Europe Air Traffic Management Research and Chapter-1.pdf (last access: 18 July 2009), 2007. Development Seminar, Baltimore, MD, 10, 2005. Naiman, A. D., Lele, S. K., Wilkerson, J. T., and Jacobson, M. Z.: Kim, B. Y., Fleming, G. G., Balasubramanian, S., Malwitz, A., Lee, Parameterization of subgrid plume dilution for use in large-scale J. J., Ruggiero, J., Waitz, I. A., Klima, K., Locke, M., Holsclaw, atmospheric simulations, Atmos. Chem. Phys., 10, 2551–2560, C. A., Morales, A., McQueen, E., and Gillette, W.: System for doi:10.5194/acp-10-2551-2010, 2010. assessing Aviation’s Global Emissions (SAGE) Version 1.5: Sys- OAG: Airlines Offer Record Number Of Seats In July, available tem Revision History, FAA, Ed., Office of Environment and En- online at: http://www.oag.com/oag/website/com/en/Press+ ergy, 21, 2005a. Room/Press+Releases+2007/Airlines+Offer+Record+Number+ Kim, B. Y., Fleming, G. G., Balasubramanian, S., Malwitz, A., Of+Seats+In+July, (last access: 27 May 2010), 2007. Lee, J. J., Ruggiero, J., Waitz, I. A., Klima, K., Stouffer, V., RIVA: The IMAGE 2.2 implementation of the SRES scenarios. A Long, D., Kostiuk, P., Locke, M., Holsclaw, C. A., Morales, A., comprehensive analysis of emissions, climate change and im- McQueen, E., and Gillette, W.: System for assessing Aviation’s pacts in the 21st century, National Institute for Public Health and Global Emissions (SAGE) Version 1.5: Technical Manual. FAA, the Environment, Bilthoven, the NetherlandsCD-ROM publica- Ed., Office of Environment and Energy, 236, 2005b. tion 481508018, 2001. Kim, B. Y., Fleming, G. G., Balasubramanian, S., Malwitz, A., Roof, C., Hansen, A., Fleming, G. G., Thrasher, T., Nguyen, A., Lee, J. J., Waitz, I. A., Klima, K., Locke, M., Holsclaw, C. A., Hall, C., Dinges, E., Bea, R., Grandi, F., Kim, B. Y., Usdrowski, Morales, A., McQueen, E., and Gillette, W.: System for assess- S., and Hollingsworth, P.: Aviation Environmental Design Tool ing Aviation’s Global Emissions (SAGE) Version 1.5: Global (AEDT) System ArchitectureAEDT-AD-01, 2007. Aviation Emissions Inventories for 2000 through 2004. FAA, Sausen, R., Isaksen, I. S. A., Grewe, V., Hauglustaine, D., Lee, D. Ed., Office of Environment and Energy, 48, 2005c. S., Myhre, G., Kohler, M. O., Pitari, G., Schumann, U., Stordal, Kim, B. Y., Fleming, G. G., Lee, J. J., Waitz, I. A., Clarke, J.-P., Bal- F., and Zerefos, C. S.: Aviation radiative forcing in 2000: An asubramanian, S., Malwitz, A., Klima, K., Locke, M., Holsclaw, update on IPCC (1999), Meteorol. Z., 14, 555–561, 2005. C. A., Maurice, L. Q., and Gupta, M. L.: System for assessing Sutkus Jr., D. J., Baughcum, S. L., and DuBois, D. P.: Scheduled Aviation’s Global Emissions (SAGE), Part 1: Model description Civil Aircract Emission Inventories for 1999: Database Devel- and inventory results, Transport. Res. Part D: Transportation and opment and Analysis, NASA-CR-2001-211216, 2001. Environment, 12D, 325–346, 2007. Volpe: Volpe National Transportation Systems Center/US DOT, Knighton, W. B., Herdon, S. C., and Miake-Lye, R. C.: Aircraft En- Enhanced Traffic Management System (ETMS), Functional De- gine Speciated Organic Gases: Speciation of Unburned Organic scription, Version 7.6VNTSC-DTS56-TMS-002, 2003. Gases in Aircraft Exhaust. Environmental Protection Agency, Waitz, I. A., Lukachko, S. P., and Lee, J. J.: Military Aviation Office of Transportation and Air Quality; Federal Aviation Ad- and the Environment: Historical Trends and Comparison to Civil ministration, Office of Environment and Energy, 2009. Aviation, J. Aircraft, 42, 329–339, 2005.

www.atmos-chem-phys.net/10/6391/2010/ Atmos. Chem. Phys., 10, 6391–6408, 2010 6408 J. T. Wilkerson et al.: Analysis of emission data from global commercial aviation

Wayson, R. L., Fleming, G. G., and Lovinelli, R.: Methodology to Wood, E. C., Herdon, S. C., Timko, M. T., Yelvington, P. E., and Estimate Particulate Matter Emissions from Certiﬁed Commer- Miake-Lye, R. C.: Speciation and Chemical Evolution of Ni- cial Aircraft Engines, J. Air Waste Manage. Assoc., 59, 91–100, trogen Oxides in Aircraft Exhaust near Airports, Environ. Sci. 2009. Technol., 42, 1884–1891, 2008. Whiteﬁeld, P. D., Lobo, P., Hagen, D. E., Timko, M. T., Miake-Lye, Zhao, Y. and Li, J.: Discrepancy of mass transport between R. C., Taylor, C., Ratliff, G., Lukachko, S., Sequeira, C., Hile- the Northern and Southern Hemispheres among the ERA-40, man, J., Waitz, I. A., Webb, S., Thrasher, T. G., Oshfeldt, M. R., NCEP/NCAR, NCEP-DOE AMIP-2, and JRA-25 reanalysis, Kaing, H. K., and Essama, S. G.: Summarizing and Interpreting Geophys. Res. Lett., 33, L20804, doi:10.1029/2006GL027287, Aircraft Gaseous and Particulate Emissions Data, Aircraft Co- 2006. operative Research Program, Report 9, Transportation Research Board, Washington, DC, 2008.

Atmos. Chem. Phys., 10, 6391–6408, 2010 www.atmos-chem-phys.net/10/6391/2010/